Thursday, November 23, 2006

AMD should bump its single core CPU clock

A 3.2GHZ single core is better than a 2GHZ dual core, and cheaper to make too.

43 Comments:

Blogger Azary Omega said...

Ok. But AMD is on brink of capacity boom, on December the 5th their going to start pushing 65nm. Each 300mm wafer will give x2.5(300mm x2 vs. 200mm x2) x1.9(90nm x2 vs. 65nm x2) x1.2(245m transistors on 90nm dual cores vs. 195m transistors on Rev F dual cores).

To me it looks like even with 50% yield they will have enough dual cores to supply 30% of the market. And when 65nm fab's yield will reach mature state - AMD will start selling their CPUs in boxes of cereal just to get read of damn things (yeah, that many)

10:02 PM, November 23, 2006  
Anonymous Anonymous said...

Yields, Doc, yields...

10:47 PM, November 23, 2006  
Anonymous Anonymous said...

"A 3.2GHZ single core is better than a 2GHZ dual core, and cheaper to make too."

...and more difficult to make too!
Further, a single 3.2 GHz core have more "hot spot" and higher power consumption than a dual 2 GHz cpu, a dual core can ever "switch off"(sort of)an unutilized core to save power.

11:32 PM, November 23, 2006  
Anonymous Anonymous said...

Intel's at it again, pressuring Via to drop processor making:

http://www.theinquirer.net/default.aspx?article=35930

11:39 PM, November 23, 2006  
Anonymous roborat said...

A 3.2GHZ single core is better than a 2GHZ dual core, and cheaper to make too.

Yes, maybe in a delusional world where increasing the target clock frequency doesn’t have a massive impact on yield. Driving your mass market processor close to its clock frequency limit will only lessen your volume significantly without getting any more profit from it. Intel has cornered this market with its lowest binning 1.8Ghz Core2Duo selling less than $200.

You and AMD should stop playing around and stop hatching ridiculous ideas with obsolete technology. Control the top and you control market pricing.

1:09 AM, November 24, 2006  
Anonymous Anonymous said...

I think you are complettly exagerating. I prefer my Single core 3700+ than one of the 3800+.

Too bad single core CPUs got removed from reviews.

2:27 AM, November 24, 2006  
Anonymous Anonymous said...

Singlecores are better only when you are not multitasking and in some older games.

Also, didn't you say some time ago that AMD65nm CPU's will be running 4GHz+?

2:34 AM, November 24, 2006  
Anonymous Anonymous said...

All that may change with multi-thread becomes more prevalent.

2:52 AM, November 24, 2006  
Anonymous Anonymous said...

i don't think that there will be many singelcore's that will run 3.2, i've triede some of these AM2 and while they are ok and decent overclockers, i have not gotten 3.2 out of any of them at standard Vcore.. but it would be nice of AMD started to make most of them at 2.6 and above, and as Cheap as possible, you could make kickass computeres at low cost, however.. Dualcore will still seem faster when you multitask.

3:18 AM, November 24, 2006  
Anonymous Anonymous said...

Why is a 3.2 GHz single core better than a 2 GHz dual-core?
Apart from that making no sense at all...
Cheaper to make - no
easier to make - hell no

5:34 AM, November 24, 2006  
Anonymous Anonymous said...

Too bad 2.8GHz singlecores cost more than 2.6GHz dualcores

http://www.newegg.com/Product/Product.asp?Item=N82E16819103558
http://www.newegg.com/Product/Product.asp?Item=N82E16819103758

2.6GHz singlecores cost more than 2.2GHz dualcores:

http://www.newegg.com/Product/Product.asp?Item=N82E16819103526
http://www.newegg.com/Product/Product.asp?Item=N82E16819103747


For some reason I doubt 3.2GHz single core CPU would be cheaper than 2GHz dualcore.

5:57 AM, November 24, 2006  
Anonymous Panther said...

Nothing better to do on thanksgiving eve?

By the way, a 3.2GHz single core processor is only good if you are a gamer or run low intensity apps.

Dual cores are the future, there is hardly a price range for a single core athlon 64 as it is right now. I can pick up an opteron 165 for 160$. You are suggesting that people buy single core processors for 150$?

A better idea would be for AMD to bump their dual-core clock speed.

6:13 AM, November 24, 2006  
Anonymous Anonymous said...

http://www.aceshardware.com/forums/read_post.jsp?id=120070892&forumid=1

Sharikou this guys say Core 2 Duo suck, and may die sooner if OC.

And I was thinking that Intel was keeping clock speeds down because they didn’t want to annihilate their 95% processor stock (P4, PD, ...).

After all there was another problem, HEAT.

7:11 AM, November 24, 2006  
Anonymous Anonymous said...

I'm sure a lot of people would agree if they sold them for $50.

7:16 AM, November 24, 2006  
Anonymous Anonymous said...

Wow doc. You're amazing.

8:48 AM, November 24, 2006  
Anonymous Anonymous said...

"After all there was another problem, HEAT."

You must be thinking lack of heat since under 70% OC my C2D is <40C on full load

10:15 AM, November 24, 2006  
Anonymous Anonymous said...

I can't believe the ignorance on this site.. YOu retards don't know shit about technology

1) 50% yields are a disaster. Anyone with 50% yields shouldn't be in the business.

2) Binsplit for higher frequency isn't hard on 65nm. If design is done right.. The issue was and is power you monkey.

11:29 AM, November 24, 2006  
Anonymous Anonymous said...

My old AMD-64 3400 (3 years old) just pasted the Vista test (Windows Vista Adviser) for business & Media. It just listed 4 drivers to download before hand (printer, graphics card, scanner and finger print reader) I checked and they are all available. I’ll buy a new machine anyway (3 years home office depreciation schedule).

Millions of consumers will be thrilled they purchased AMD 64 and won’t need a new machine. At the time of purchase the box wasn’t anymore than a 32-bit P3/P4 machine.

To check your machine;
http://www.microsoft.com/windowsvista/getready/upgradeadvisor/default.mspx

12:02 PM, November 24, 2006  
Anonymous Anonymous said...

Anónimo dijo...

I can't believe the ignorance on this site.. YOu retards don't know shit about technology

1) 50% yields are a disaster. Anyone with 50% yields shouldn't be in the business.

2) Binsplit for higher frequency isn't hard on 65nm. If design is done right.. The issue was and is power you monkey.

11:29 AM, November 24, 2006

tell that to SONY SIRE!!
I doubt conroe started with more than 20% yield..

2:16 PM, November 24, 2006  
Anonymous Anonymous said...

Wait - wasn't that an old intel tactic? "Higher Clock Speeds = better processor?"

So it looks like you took Intels previous idiotic theory with you when they fired you, huh?

8:42 PM, November 24, 2006  
Anonymous girly bates said...

Maybe instead of "4x4", AMD should put 8 sockets on a board, and go for "1x8".

They could call it OCTOPEDE64.

You better send this intelligence to the big brains at AMD right away.

6:24 AM, November 25, 2006  
Anonymous Anonymous said...

“AMD Server Revenue Up 79.7% Year on Year”

Do you Intel-ies see the trend? Can you see the Woodcrest isn’t helping your cause? Even with all the Intel hype your still loosing market share. Do you know why?

6:39 AM, November 25, 2006  
Anonymous Anonymous said...

“I can't believe the ignorance on this site.. YOu retards don't know shit about technology”

“1) 50% yields are a disaster. Anyone with 50% yields shouldn't be in the business.”



Please read and comprehend. The blogger stated “even with” which is the article before the preposition (50% yields). Which means EVEN WITH a 50% yield you could supply 30% of the market. It doesn’t mean they only have a 50% yield.

That’s how I see it and I flunked English.

7:31 AM, November 25, 2006  
Anonymous edward said...

"If design is done right.. The issue was and is power you monkey."

Take a look at P4EE. Even if the design is done NOT right, power doesn't seem to be an issue (for "EE"). Or does it?

8:13 AM, November 25, 2006  
Anonymous Anonymous said...

“2) Binsplit for higher frequency isn't hard on 65nm. If design is done right.. The issue was and is power you monkey.”



65nm is about power, however due to large cache dependency in Intel’s prehistoric design Intel needs 65nm. It takes a lot of power to turn on/off all those transistors in large cache, which is where Intel gets its speed.

65nm helps the power hungry design. I’m not an historian but I believe Columbus brought this design over on the Mayflower. The design was by monkeys for monkeys.

8:26 AM, November 25, 2006  
Anonymous Anonymous said...

“Anonymous said...
I can't believe the ignorance on this site.. YOu retards don't know shit about technology”


Oh my………so much pain. I’m going to recommend you should take a time out. Better yet, go stand in the corner.

9:23 AM, November 25, 2006  
Anonymous Anonymous said...

“ Intel's at it again, pressuring Via to drop processor making:’


The only way Intel can get market share is push Via around.

Pathetic!

9:27 AM, November 25, 2006  
Anonymous Anonymous said...

"65nm is about power, however due to large cache dependency in Intel’s prehistoric design Intel needs 65nm. It takes a lot of power to turn on/off all those transistors in large cache, which is where Intel gets its speed"

Actually the cache is the coolest part of the chip.

10:15 AM, November 25, 2006  
Anonymous Anonymous said...

Even with all the Intel hype your still loosing market share. Do you know why?

Intel gained x86 market share sequentially for the first time in 4 years and Intel-branded processors maintained 80.2% of all x86 server spending in the quarter.

Come again?

11:46 AM, November 25, 2006  
Anonymous Anonymous said...

I’m not an historian but I believe Columbus brought this design over on the Mayflower. The design was by Intel for Intel.

12:37 PM, November 25, 2006  
Anonymous Anonymous said...

Why do Intel-ies defend what Intel its self is ashamed of?

On the eve of Conroe release they also made this announcement “were going to release a new architecture every two years”.

TRANSLATION; we know this is a tweaked P3 with large cache that still uses FSB.

At the time Intel was being brutally hammered about the power to watt performance issues and they knew they would be buying some time beating AMD to 65nm thus making their CPU’s temporarily expectable. Intel was embarrassed because it was widely known they were caught sleeping and were releasing a newly package P3.

Further proof was shortly after they released new commercials that started using the slogan “Worlds Best Processors”. They chose the title them selves, how convenient.
TRANSLATION; This architecture will be replace as soon as possible, but until then our new slogan will keep our shareholders happy and the public will buy anything.

Intel figured there architecture would last for years, how else can you explain the huge overstock.

1:19 PM, November 25, 2006  
Anonymous Anonymous said...

You Ineliots still don't get it.Conroe is competing against a 3-year old K8 design and beats it -JUST. Once the K8L comes out, Conroe will be consigned to the dustbin, where it will join HeatBurst et al.
Furthermore, if Conroe is so good, why are Intel going to bring out a new architecture every 2 years??

6:28 PM, November 25, 2006  
Anonymous Anonymous said...

The AMD fanboys like claiming that the Core 2 CPU is just "a repackaged Pentium 3". Does this mean that AMD's best processors are beaten by over 30% by Pentium 3s!? That sounds even more pathetic for AMD. As the results in Flash benchmarks at gamepc.com clearly show:- AMD FX 62 is slower than the cheapest Core 2 E6300.

8:27 PM, November 25, 2006  
Anonymous Anonymous said...

Here's an interesting take.

http://voodoopc.blogspot.com/2006/11/intel-from-winnie-pooh-to-red-bull.html

9:52 PM, November 25, 2006  
Anonymous Anonymous said...

“Intel gained x86 market share sequentially for the first time in 4 years and Intel-branded processors maintained 80.2% of all x86 server spending in the quarter.”


Now this is interesting, AMD claims 79.7% increase year over year in server and Intel claims it has maintained its market share in x 86 servers.

4:44 AM, November 26, 2006  
Anonymous Anonymous said...

"Furthermore, if Conroe is so good, why are Intel going to bring out a new architecture every 2 years??"

Because they can create a better architecture after every two years?

Basically Intel's plan is this:

1) Create a new architecture on current processing design with N cores
2) slap two of the cores to one die for N+N cores (a'la Kentsfield)
3) move to new manufacturing technology
4) create a replacement for N+N CPU with 2N CPU (a'la native quadcore that comes after Kentsfield)
5) repeat

A new core comes out in around step 4

"Now this is interesting, AMD claims 79.7% increase year over year in server and Intel claims it has maintained its market share in x 86 servers."

It's simple: server market increased as a whole. Intel reported it kept the market share it had before, AMD reported it sold more than last year. Simple.

9:05 AM, November 26, 2006  
Blogger TheKhalif said...

Now this is interesting, AMD claims 79.7% increase year over year in server and Intel claims it has maintained its market share in x 86 servers.


That is revenue not shipments. That basically means that the server market is still growing and that AMD is making more per sale now.

9:16 AM, November 26, 2006  
Blogger Sharikou, Ph. D said...

Furthermore, if Conroe is so good, why are Intel going to bring out a new architecture every 2 years??


Intel won't have a chance to bring out another architecture. It will have BKed before that. Everyone knows FSB architecture is obsolete. K8L will be a killer.

6:14 PM, November 26, 2006  
Anonymous Anonymous said...

Everyone knows FSB architecture is obsolete

And still it is outperforming Athlon 64 in almost all consumer benchmarks.

3:18 AM, November 27, 2006  
Anonymous Anonymous said...

"Everyone knows FSB architecture is obsolete. K8L will be a killer."

I can't wait to see what you will say when 45nm native-quadcore from Intel goes head-to-head against AMD quadcore :)

7:39 AM, November 27, 2006  
Anonymous Anonymous said...

Anonymous said...

"And still it is outperforming Athlon 64 in almost all consumer benchmarks."

Benchmarks like the toy benchmark Super Pi? Against a 3 year old processor? Oooooh yea, real impressive.

In Intel's credit, it's a somewhat known fact Core 2 does better in SIMD (SSE) functions,(no surprise, P4 was the same way) however, in FP functions shows a different story and in 64-bit applications/OS an even more different story. While the gap is still there it's down to a mere %5-%10 deficit at best, hardly noticable in real world experiences IMHO.

As things are progressing, it's becoming obvious that the bus is becoming more and more vital to the overall system performance and less reliance on the processor for sheer performance(especially since the move to multi-core processors) as everything is basically "tied" to it concerning data-bandwidth. HT's point to point serial bus is still light years ahead of the Intel GTL bus in this area, especially with it's ability to scale in performance as the processor speeds/performance goes up. You can have the fastest processor on the planet, but if you can't feed it data fast enough, it's performance is ultimately going to be hindered. Intel can only "band-aid" the old GTL bus for so long to keep performance.(hence the eventual move to CSI/IMC)

Looking forward though, in about 6-7 months it won't matter though, Intel's Core 2 will more than meet it's match with K8L.

6:59 AM, November 28, 2006  
Anonymous Anonymous said...

" Against a 3 year old processor? Oooooh yea, real impressive. "

Yes, against 3 year old benchmark with a CPU that simply a bit improved >6y old architecture.

3:22 AM, November 29, 2006  
Anonymous Anonymous said...

Isn't it funny how Intel's new design based around an older architecture soundly beats AMD's newer designs? (newer for AMD being 3 years .vs Intels 6 year old architecture)

Isn't it funnier how the benches that have hit the streets today (both real world apps and synthetic benches) show the same architecture from AMD on a quad platform (4x4) getting soundly beat by Intels "older" architecture tweaked for the quad Core 2?

4x4 is a failure. By the time AMD comes out with K8L Intel will already have a counter to that as well. I doubt AMD will go under, and I think the competition is great, but the Pretender's ludicrous assertations that Intel will go under 2Q08 are farts in the wind. Just a disgruntled ex Intel employee who got fired for being a dumbass.

11:16 AM, November 30, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home